可能的提示信息如下:
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ @ WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED! @ @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ IT IS POSSIBLE THAT SOMEONE IS DOING SOMETHING NASTY! Someone could be eavesdropping on you right now (man-in-the-middle attack)! It is also possible that the RSA host key has just been changed. The fingerprint for the RSA key sent by the remote host is 23:00:20:83:de:02:95:f1:e3:34:be:57:3f:cf:2c:e7. Please contact your system administrator. Add correct host key in /home/xahria/.ssh/known_hosts to get rid of this message. Offending key in /home/xahria/.ssh/known_hosts:8 RSA host key for localhost has changed and you have requested strict checking. Host key verification failed. |
下面简单讲一下这个问题的原理和比较长久的解决方案。
用OpenSSH的人都知ssh会把你每个你访问过计算机的公钥(public key)都记录在~/.ssh/known_hosts。当下次访问相同计算机时，OpenSSH会核对公钥。如果公钥不同，OpenSSH会发出警告，避免你受到DNS Hijack之类的攻击。
SSH对主机的public_key的检查等级是根据
StrictHostKeyChecking变量来配置的。默认情况下，
StrictHostKeyChecking=ask。简单所下它的三种配置值：
1.
StrictHostKeyChecking=no
#最不安全的级别，当然也没有那么多烦人的提示了，相对安全的内网测试时建议使用。如果连接server的key在本地不存在，那么就自动添加到文件中（默认是known_hosts），并且给出一个警告。
2.
StrictHostKeyChecking=ask #默认的级别，就是出现刚才的提示了。如果连接和key不匹配，给出提示，并拒绝登录。
3.
StrictHostKeyChecking=yes #最安全的级别，如果连接与key不匹配，就拒绝连接，不会提示详细信息。
对于我来说，在内网的进行的一些测试，为了方便，选择最低的安全级别。在.ssh/config（或者/etc/ssh/ssh_config）中配置：
| StrictHostKeyChecking no UserKnownHostsFile /dev/null
Solved : Error displaying connection information: No valid active connections found! in Lucid28082010Beberapa hari yang lalu network-manager tidak muncul/tampil di applet pada Ubuntu Lucid stelah update. Isi file : Ubah menjadi : Simpan dan keluar, lalu restart service network-manager dan network.
Selesai smoga bermanfaat. |
（注：这里为了简便，将knownhostfile设为/dev/null，就不保存在known_hosts中了）
http://www.debian-administration.org/articles/476
There are times when people make mistakes, and manage to do crazy things to working systems. It wasn't so long ago that a hasty deletion caused me all kinds of problems. Recently I read of an unfortunate sysadmin who managed to recursively change permissions on their root filesystem - and here is my attempted solution.
The problematic command was:
(none):~# chmod -R 777 /
Here the entire filesystem, and all mounted partitions, were modified to be world-writable. One immediate consequence of this is that ssh logins fail:
skx@lappy:~$ ssh 192.168.1.199 -l root ssh_exchange_identification: Connection closed by remote host
Looking in /var/log/auth.log upon the host we can see the cause of this failure:
error: Could not load host key: /etc/ssh/ssh_host_rsa_key error: @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ error: @ WARNING: UNPROTECTED PRIVATE KEY FILE! @ error: @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ error: Permissions 0777 for '/etc/ssh/ssh_host_dsa_key' are too open. error: It is recommended that your private key files are NOT accessible by others. error: This private key will be ignored. error: bad permissions: ignore key: /etc/ssh/ssh_host_dsa_key error: Could not load host key: /etc/ssh/ssh_host_dsa_key fatal: /var/run/sshd must be owned by root and not group or world-writable.
These errors are fixable fairly easily since we still have an open root session:
(none):~# chmod 600 /var/run/sshd (none):~# chmod 600 /etc/ssh/ssh_host_*
Since we have root access we can now fix things up, and the availability of remote ssh access is a nice bonus. However these remote logins cannot become root themselves just yet - because su and sudo have lost their setuid flags:
skx@(none):/$ su Password: setgid: Operation not permitted
We can fix this fairly easily, again using our existing root login:
(none):~# chmod 4755 /bin/su /usr/bin/sudo
Now we can login remotely and become root. So the system is 90% fixed!
But what if we didn't have root access? What if we could only login via a normal user account, via the console perhaps?
Lets try that now.
(none):~# chmod -R 777 / (none):~# exit
Remote logins via ssh will have become broken again - how can we fix things now? Suddenly we can't fixup permissions as we're not root, and we can't become root since the binaries are no longer setuid.
It seems that our first job is to gain root access. Whilst there are many things we could try (such as emptying the password fields in /etc/passwd - possible now all files are world-writable) first solution to occur to me was to abuse a daemon which is already running and which has permissions to do root-type jobs. In a word: cron.
Since the system has various php4 packages installed we can modify the cronjob which has been installed:
echo "*/1 * * * * root cp /bin/sh /tmp/sh && chmod 4755 /tmp/sh" > /etc/cron.d/php4
Waiting a minute finds us a setuid shell in /tmp and we can use that to become root and fix things as we did previously:
skx@uml:~$ /tmp/sh sh-3.1# id uid=1000(skx) gid=1000(skx) euid=0(root) groups=1000(skx)
Having writable files will allow a lot of different attacks against a system, cron is just the most obvious. The important thing is to not panic, and do anything hasty such as rebooting or exiting your root shell!
[ Send Message | View Steve's Scratchpad | View Weblogs ]
I guess it did just tail off a little, but the hard parts were done. Once you can login as root remotely then you're easily able to do the site-specific fixups from the comfort of your desk.
If it were me I'd just restore from backups which would incidentally fix the permissions!
Still if you have another working system I guess the obvious thing to do is to write a simple script to output a file containing the path + permissions of every file on the system. Then another to read in that file and apply the fixups.
I remember doing that once in the past. Something like this simple save.pl script would do the job:
#!/usr/bin/perl -w
use strict;
use File::Find;
find(\&wanted, "/");
sub wanted
{
my $file = $File::Find::name;
my $mode = (stat($file))[2];
printf "%04o, %s\n", $mode & 077777, $file;
}
This will give output such as this:
40755, / 0600, /.bash_history 0600, /.viminfo 0600, /.lesshst 0600, /initrd.img.old 0644, /vmlinuz.old 40755, /lost+found 40755, /home
The script to restore those permissions is left as an excercise ;)
[ Parent | Reply to this comment ]
getfacl -R / > /home/backups/acl-`date +%d-%m-%Y`.txt
with setfacl you can recover this :)
[ Parent | Reply to this comment ]
Kind regards
Wolfgang Karall
--
Debian GNU/Linux on an IBM Thinkpad T43p
[ Parent | Reply to this comment ]
[ Parent | Reply to this comment ]
> ldd /bin/sh
linux-gate.so.1 => (0xffffe000)
libncurses.so.5 => /lib/libncurses.so.5 (0xb7eaf000)
libdl.so.2 => /lib/tls/i686/cmov/libdl.so.2 (0xb7eaa000)
libc.so.6 => /lib/tls/i686/cmov/libc.so.6 (0xb7d76000)
/lib/ld-linux.so.2 (0xb7f1e000)
So if you screw up your dyn libraries somehow (move them aside while trying to localize possible intrusion detected by the simple md5 sum check... don't ask) -- then you are doomed to fail starting up any new application which relies on dyn libraries, including /bin/sh.
Whenever I got into such situation, I did very good thing -- I didn't panic, took beer and a cigarette and just relaxed to become sober in my actions ;-) (to don't say that I was ~6kmiles away from the system, so I just had ssh/VNC access... where obviousely no new connections could be setup)
Lyrics aside, keep some staticly compiled shell readily available:
sash - Stand-alone shell zsh-static - A shell with lots of features (static link) bash-static - The GNU Bourne Again SHell (static version)
sash installation (may be others as well) optionally does a handy thing - creates sashroot user which uses sash as his shell ;-) so if system is %%ed you still have a chance to login remotely (Surely you better you need to restrict access to that account...)
[ Parent | Reply to this comment ]
Boot from rescue CD, pick unused console, "chroot /mnt/rootfs ; passwd".
Root doesn't have to give the old password to set a new one.
I'm not sure one should rely on having insecure temporary files around the place to become root.
I'm not sure leaving a system with warped file permissions up is a good idea either, but hey Linux and Unix are remarkably robust to such cock-ups.
Perhaps the next article should be on the U-Kuang program, and how to spot when you've left stupid file permissions laying around.
[ Parent | Reply to this comment ]
What I'd like to know is once you have root access is there a simple way to get apt/aptitude to refresh all the permissions in /usr /sbin etc?
The same question applies to the situation where you accedently delete part of /usr. Is there a way to get apt to verify every install package?
<disclaimer> I probably should just use google to answer this question, but if you could post and answer here it would be a good resource for people.<disclaimer>
Elivs
[ Parent | Reply to this comment ]
[ Send Message | View Steve's Scratchpad | View Weblogs ]
Not directly; although you could run "apt-get install --reinstall [package1] [package2] .." to do it.
Failing that you could install the acl package and use getfacl + setfacl to create/restore from a backup as suggested by mvanbaak already.
In the case of missing files then the "--reinstall" trick might work, but the obvious answer is "restore from backups". If you don't currently have backups you'll have learnt how important they are for the future ;)
[ Parent | Reply to this comment ]
[ Parent | Reply to this comment ]
[ Parent | Reply to this comment ]
command on the root filesystem.
I wanted to just put the modes back, by duplicating them from a similar
desktop system ... just the items that matched name and location.
I couldn't figure out how to do that using rsync at the time, not sure if
it was feasible.So I wrote a shell script that accomplished it, I don't remember if it was based on cpio or something else. Now I'm thinking that I may have written some perl code that traversed the system, operating on each directory that existed on both sides.Anyhow, I think that it's possible to use rsync to derive just the modes, if you have a second system that you want to copy modes from (assuming that you've fixed the broken system enough to make the basic commands and networking run, and you just want to clean up the rest of the system areas). Larry
[ Parent | Reply to this comment ]
apt-get --reinstall install $(dpkg -S /usr/bin/* | awk ' { print $1 } ' | sort -u | sed -e 's/:$//')
Should do the package reinstall.
I had a remote login with ssh-keygen ability from my master server. Otherwise there would have been no ability to change to root user.
regards
bctechnzl
[ Parent | Reply to this comment ]
The command issued was to allow group read / write permissions. So this did really destroy my system, but not so much that I could easily regain root access via the methods mentioned in the post (I had no root shell).
Someone on another forum mentioned pkexec. Apparently pkexec is not that choosy about file permissions and will let me get a root shell as long as I am logged in as a user in the admin group. This allowed me to get root access with the running ssh session.
Thanks for the original post, it helped!
[ Parent | Reply to this comment ]
find / -type f -print0 2>/dev/null |xargs -0 ls -al |awk -F " " {'print $9" "$3" "$4" "$1'} >all_perms_files
+ perl script(which make bash script for recovery permissions).
5601
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
[ Send Message ]
Kind regards
Wolfgang Karall
--
Debian GNU/Linux on an IBM Thinkpad T43p
[ Parent | Reply to this comment ]